Added _Py_HOT_FUNCTION macro

third_party/python/Include/pyport.h

@@ -235,6 +235,28 @@ typedef int Py_ssize_clean_t;
 #define Py_DEPRECATED(VERSION_UNUSED)
 #endif
 
+/* _Py_HOT_FUNCTION
+ * The hot attribute on a function is used to inform the compiler that the
+ * function is a hot spot of the compiled program. The function is optimized
+ * more aggressively and on many target it is placed into special subsection of
+ * the text section so all hot functions appears close together improving
+ * locality.
+ *
+ * Usage:
+ *    int _Py_HOT_FUNCTION x(void) { return 3; }
+ *
+ * Issue #28618: This attribute must not be abused, otherwise it can have a
+ * negative effect on performance. Only the functions were Python spend most of
+ * its time must use it. Use a profiler when running performance benchmark
+ * suite to find these functions.
+ */
+#if !IsModeDbg() && defined(__GNUC__) \
+    && ((__GNUC__ >= 5) || (__GNUC__ == 4) && (__GNUC_MINOR__ >= 3))
+#define _Py_HOT_FUNCTION __attribute__((hot))
+#else
+#define _Py_HOT_FUNCTION
+#endif
+
 #define RTYPE RTYPE
 #ifdef __cplusplus
 #define PyMODINIT_FUNC extern "C" PyObject *

third_party/python/Objects/call.c

@@ -279,7 +279,7 @@ PyObject_Call(PyObject *callable, PyObject *args, PyObject *kwargs)
 
 /* --- PyFunction call functions ---------------------------------- */
 
-static PyObject* 
+static PyObject* _Py_HOT_FUNCTION
 function_code_fastcall(PyCodeObject *co, PyObject **args, Py_ssize_t nargs,
                        PyObject *globals)
 {

third_party/python/Objects/dictobject.c

@@ -748,7 +748,7 @@ the <dummy> value.
 For both, when the key isn't found a DKIX_EMPTY is returned. hashpos returns
 where the key index should be inserted.
 */
-static Py_ssize_t
+static Py_ssize_t _Py_HOT_FUNCTION
 lookdict(PyDictObject *mp, PyObject *key,
          Py_hash_t hash, PyObject ***value_addr, Py_ssize_t *hashpos)
 {
@@ -862,7 +862,7 @@ top:
 }
 
 /* Specialized version for string-only keys */
-static Py_ssize_t
+static Py_ssize_t _Py_HOT_FUNCTION
 lookdict_unicode(PyDictObject *mp, PyObject *key,
                  Py_hash_t hash, PyObject ***value_addr, Py_ssize_t *hashpos)
 {
@@ -936,7 +936,7 @@ lookdict_unicode(PyDictObject *mp, PyObject *key,
 
 /* Faster version of lookdict_unicode when it is known that no <dummy> keys
  * will be present. */
-static Py_ssize_t
+static Py_ssize_t _Py_HOT_FUNCTION
 lookdict_unicode_nodummy(PyDictObject *restrict mp, PyObject *restrict key,
                          Py_hash_t hash, PyObject ***value_addr,
                          Py_ssize_t *hashpos)
@@ -1003,7 +1003,7 @@ lookdict_unicode_nodummy(PyDictObject *restrict mp, PyObject *restrict key,
  * Split tables only contain unicode keys and no dummy keys,
  * so algorithm is the same as lookdict_unicode_nodummy.
  */
-static Py_ssize_t
+static Py_ssize_t _Py_HOT_FUNCTION
 lookdict_split(PyDictObject *mp, PyObject *key,
                Py_hash_t hash, PyObject ***value_addr, Py_ssize_t *hashpos)
 {

third_party/python/Objects/frameobject.c

@@ -461,7 +461,7 @@ static int numfree;
 static PyFrameObject *free_list;
 #define PyFrame_MAXFREELIST 200
 
-static void
+static void _Py_HOT_FUNCTION
 frame_dealloc(PyFrameObject *restrict f)
 {
     PyObject **p, **valuestack;
@@ -660,7 +660,7 @@ int _PyFrame_Init()
     return 1;
 }
 
-PyFrameObject*
+PyFrameObject* _Py_HOT_FUNCTION
 _PyFrame_New_NoTrack(PyThreadState *tstate, PyCodeObject *code,
                      PyObject *globals, PyObject *locals)
 {

third_party/python/Python/ceval.c

@@ -732,7 +732,7 @@ PyObject *
     return tstate->interp->eval_frame(f, throwflag);
 }
 
-PyObject *
+PyObject * _Py_HOT_FUNCTION
 _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
 {
 #ifdef DXPAIRS
@@ -4813,7 +4813,7 @@ if (tstate->use_tracing && tstate->c_profilefunc) { \
     x = call; \
     }
 
-static PyObject *
+forceinline PyObject * _Py_HOT_FUNCTION
 call_function(PyObject ***pp_stack, Py_ssize_t oparg, PyObject *kwnames)
 {
     PyObject **pfunc = (*pp_stack) - oparg - 1;

third_party/python/Python/errors.c

@@ -153,7 +153,7 @@ PyErr_SetString(PyObject *exception, const char *string)
     Py_XDECREF(value);
 }
 
-PyObject *
+PyObject * _Py_HOT_FUNCTION
 PyErr_Occurred(void)
 {
     PyThreadState *tstate = _PyThreadState_UncheckedGet();